Critical Appraisal Worksheet for Studies about Prognosis

Dalton: Risk for Hospitalization With Depression After a Cancer Diagnosis: A Nationwide, Population-
Based Study of Cancer Patients in Denmark From 1973 to 2003
Group 1 – 1 to 2
Group 2 – 3 to 4
Group 3 – 5 to 6
Group 4 – 7 to 8
Group 5 – 9 to 11
All – question 12

Based on: Fineout-Overholt, E., & Melynk, B. M. (2008). Evaluation of studies of prognosis. In N. Cullum, D. Ciliska,
R. B. Haynes, S. Marks (Eds.), Evidence-based nursing: An introduction (pp. 168-177). Blackwell
Publishing Ltd.

Questions Comments

Are the Results Valid?

*Group 1

P adults over 15 years

1. Did the study ask a clearly focused
question? E cancer survivors
▪ Consider if the question in “focused” C cancer-free
terms of PICO/PECO O hospitalization for depression
Study question was somewhat focused.
Focused-specified follow up, had comparison
group, unlike some other cohort studies
2. What type of design is used in this *Group 1
study? Was this the best design to Case-control was used cohort was also used
answer the question that the study Case-control portion-retrospective observation
was asking?
Comparison group relative risk was set as 1, the
▪ Consider appropriate use of case-
normal
control or cohort studies
Cohort looks at registry, see Question 3.
This was appropriate because:
- Cancer ‘rare’
- Data already collected
RTC unethical; can’t randomize someone to have
cancer
3. Was there a representative and *Group 2
well-defined sample of patients at Yes, representative
a similar point in the course of the - Civil Registration System linked with Danish
disease?
Cancer Registry born between 1973-1988
▪ How was the sample selected? - Cancer diagnosis 1973-2003
▪ Are the participants at a similar, - Participants selected for age, over 15 years
clearly described point in the - Cancers grouped by major etiologic factors
disease process? o type of disease
- Hospitalization of depression linked
o No fees, no private admission
-
4. Was follow-up sufficiently long and *Group 2
complete? Length
▪ Was follow-up sufficiently long - Average follow up 4.2 years
enough to observed the anticipated
- Highest risk for admission from depression
outcome?
for 1 year after Dx, can remain increased
▪ For cohort studies what proportion
for up to 10 years
of the participants in each group
- Up to 30 years, sufficiently long
that did not complete the study?
Drop-out Rates
Were the percentages similar? Are
the drop-out rates acceptable?

- consider emigration and disappearance as

drop-out; rate 2.2%
- Maximum acceptable drop-out rate 20%
5. Were objective and unbiased *Group 3
outcome criteria used? - National registries are very unbiased
▪ Were the outcomes clearly defined Yes, the outcome was clearly defined at the beginning of the
at the beginning of the study? study (hospitalization of depression)
▪ Were the staff/study personnel Yes the outcome was objective + unbiased criteria were
assessing the outcome(s) blinded to used: ICD codes
prognostic factors? Note for - ICD 8 + 10 codes for psychiatric diagnoses were
used to identify depression (ICD codes are
objective measures such as death,
internationally accepted codes representing
blinding may not be necessary. diagnoses)
- Researchers used data collected in the Danish
Psychiatric Central Register to determine whether
hx for depression occured
Yes, those assessing the outcome were blinded to prognostic
factors. Because data was collected by HCP (thru the Danish
Psychiatric Central Register, - the physicians, nurses, staff
entering, date of admission/discharge, etc.) prior to the
study. (i.e. the people collecting the data had no idea that
someone was going to a study on it (done retrospectively)

6. Did the analysis adjust for *Group 3

important prognostic factors? Separate analyses were conducted to compare different
▪ Were different groups of patients groups
identified based on prognostic - Male vs female
- Cancer types, including:
factors? Were these factors taken
o smoking-related cancers
into account (adjusted for) in the o alcohol-related cancers
analysis? o virus and immune-related cancers
o hormone related cancers
o colorectal cancers
o brain cancer
Did adjust for age, gender, extent of disease (Stage 1 vs 4)
etc. smoking, alcohol, colorectal, brain
Did not adjust for everything (recurrence, pain, other cancer
symptoms, etc.) - but did for the most important ones
7. Based on the strengths and *Group 4
limitations identified throughout Moderately strong
the appraisal, rate the study as Strengths
methodologically strong,
- large, population based data set
moderate, or weak.
- almost all patients were followed up with
- differentiated groups of related cancers
- used psychiatric assessments
- services available not limited by economic
factors: accessed by control also accessed
by intervention
Limitations

- other prognostic factors (recurrence, pain,

other cancer symptoms) not adjusted for
- only HOSPITALIZED depression considered
What are the results? Risk for depression in the first year after cancer
diagnosis increased. For most specific cancers, the
risk decreased but was still in significant excess risk
in years following the diagnosis. Exceptions where
the risk increased throughout the study period
was for those surviving hormone-related cancers,
smoking-related cancers for women, and
virus/immune-related cancers for men.
8. Consider the outcome of *Group 4
hospitalizations for depression for Good/bad outcome
“All” disorders and “Hormone - Hospitalization for depression is a bad
related cancers” in women only if
outcome; thus it is analysed using RR
your current care scenario is “I’m
Analysis
Too Young to Have Breast Cancer”
and “Colorectal cancer” in men - RR is used to analyse dichotomous
only if your current care scenario is outcomes (hospitalization for depression,
“When A Cure Isn’t Possible.” no hospitalization for depression)
▪ Is the outcome a “good” outcome or Colorectal cancer in men/When A Cure Isn’t
“bad” outcome? Possible
▪ Was the outcome analysed as a
continuous or dichotomous
outcome (or both)?

- A confidence interval of 1.67 to 2.75 for

first year of follow up is statistically
significant and a bad outcome since it is
much greater than that of no diagnosis
- The relative risk of 2.14 means men with
colorectal cancer during the first year of
follow-up are 2 times more likely to be
hospitalized for depression with that
diagnosis and timeframe
9. How likely were the outcomes over *Group 5
time? - “More recent studies of breast cancer
patients reported that 17% to 28% scored
16 or more on the Center for Epidemiologic
Studies Depression Scale, indicating an
increased risk of clinical depression at the
end of primary treatment; and in a sample
of breast cancer patients up to 5 years
after diagnosis, 9% had major depressive
disorder measured by the Structured
Clinical Interview for DSM-IV.”
- RR drops from first year to 1-4 to 5-9 to
greater than 10, decreases
10. Is the estimate of likelihood *Group 5
statistically significant? Statistically significant because the CI does not
▪ Consider the confidence interval (CI) contain 1-- D1CHOTOMOUS
first, then the p-value All types cancer more likely to be hospitalized for
 depression (except 10 years follow up for women)

11. Is the estimate of the likelihood *Group 5

precise? Best case scenario: lower end of CI, worst case:
▪ Consider the confidence interval (CI) higher end of CI
▪ Consider if you the results would
lead to selecting or avoiding at the The estimate of the likelihood is precise because
lower end of the CI? differences between Cis are quite narrow, which
adds precision regarding the likelihood of the
outcome in a population.

Since RRs are close to the high end of the CI (bad

outcome), results may lead to opting to intervene
during 1st year of follow up.
Can I Apply the Results?
12. How would I apply this evidence in *All groups
nursing practice?
▪ Is the population, exposure,
outcome(s) in the study similar to
that in the scenario?
▪ Would the results lead to selecting
or avoiding therapy?
▪ Are the results useful for reassuring
or counselling patients?
▪ Consider the Model for Evidence
Based Clinical Decisions

beck’s scale- measuring the depression level

BEFORE hospitalization
study - cancer patients higher risk for depression;
specific diagnosis, better informed care